Dropping based contention resolution for service differentiation to provide qo s
Upcoming SlideShare
Loading in...5
×
 

Dropping based contention resolution for service differentiation to provide qo s

on

  • 415 views

 

Statistics

Views

Total Views
415
Views on SlideShare
415
Embed Views
0

Actions

Likes
0
Downloads
2
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Dropping based contention resolution for service differentiation to provide qo s Dropping based contention resolution for service differentiation to provide qo s Document Transcript

  • INTERNATIONALComputer EngineeringCOMPUTER ENGINEERING International Journal of JOURNAL OF and Technology (IJCET), ISSN 0976- 6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEME & TECHNOLOGY (IJCET)ISSN 0976 – 6367(Print)ISSN 0976 – 6375(Online)Volume 4, Issue 1, January- February (2013), pp. 218-228 IJCET© IAEME: www.iaeme.com/ijcet.aspJournal Impact Factor (2012): 3.9580 (Calculated by GISI) ©IAEMEwww.jifactor.com DROPPING BASED CONTENTION RESOLUTION FOR SERVICE DIFFERENTIATION TO PROVIDE QoS IN WDM OBS NETWORKS Prof. J. R. Pathan1, Prof. A. R. Teke2, Prof. M. A. Parjane3, Prof. P.S. Togrikar4 1, 2 Department of Computer Engineering, S. B. Patil College of Engineering, Indapur, Pune (MS), India 3, 4 Department of E & TC Engineering, S. B. Patil College of Engineering, Indapur, Pune (MS), India pathan.pathan@gmail.com1, aish_art@rediffmail.com2, mparjane@gmail.com3, pradeep.togrikar@gmail.com4 ABSTRACT In the recent years, the bandwidth demand on the Internet is increasing phenomenally, due to rising of multimedia applications and real time applications. WDM technology enables to respond to the explosive growth of Internet traffic by exploiting huge bandwidth of the fiber. Optical burst switching (OBS) in WDM is a promising technology to use in the next generation optical network. One of the challenges in OBS based network is wavelength contention to provide quality of service (QoS). QoS helps to provide priorities to different applications depending on their bandwidth requirement. In case of insufficient bandwidth, QoS also helps to guarantee the burst dropping probability. In this paper, an effective approach is proposed to provide QoS in OBS networks. The primary aim of the proposed approach is to reduce loss of high delay sensitive data. High delay sensitive data is allowed to lease the channels which are allocated to the data having low priority. We put counters at each core node to measure number of low priority and high priority burst drop. If value reaches to its threshold then respective burst is not allowed to drop. Simulations are conducted to evaluate the performance of proposed algorithm in terms of burst dropping probability. The performance of proposed algorithm is then compared with LAUC-VF algorithm. The dropping probability behaviour of high priority and low priority data bursts with increasing load of high delay sensitive data burst is studied. Keywords: WDM, Optical burst switching (OBS), Quality of service (QoS), Service differentiation, Burst dropping probability. 218
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEMEI. INTRODUCTION In WDM system, each fiber carries multiple communication channels with eachchannel operating on a different wavelength. Since, it is an attractive optical multiplexingtechnique that allows better exploitation of the fiber capacity and reducing the cost ofcore network. Hence, WDM has received much more attention as a promising approachto building the next generation Internet. There are several optical network switchingparadigms, namely optical circuit switching (OCS), optical packet switching (OPS), andoptical burst switching (OBS). Of all these paradigms OBS combines the best of circuitswitching and packet switching while avoiding their shortcomings [5]. Due to thestatistical multiplexing in OBS, an improvement is observed in efficiency and scalabilityas compared with OCS. In OBS, the bandwidth utilization is more efficient than OPS. Inan OBS based network, a burst consisting of multiple IP packets is switched through thenetwork all optically [7]. The control packet (header of a burst) is transmitted on aseparate wavelength ahead of the transmission of burst payload to ensure sufficient timefor header processing [5]. Due to the one-way reservation scheme, burst loss may occur inan OBS network because the control may not succeed in reserving resources at some ofthe intermediate OBS core nodes [9]. Although OBS provides better solution over OCS and OPS, several issues likeburst assembly, signalling schemes, contention resolution, burst scheduling, and qualityof service (QoS) need to be consider. Contention occurs when more than one burstcontends for the same resource at the same time. In an OBS based networks, contentionamong bursts can be resolved using deflection, dropping, and pre-emption [7]. It is alsopossible to break the incoming burst into multiple segments, and each segment can thenbe deflected, dropped and pre-empted [7]. In this paper, we are using dropping basedapproach for contention resolution. Another challenging issue in OBS based networks is support to QoS. Thebandwidth requirement for the multimedia and real time applications is increasingphenomenally. For such applications, there should be sufficient bandwidth available toguarantee dropping probability. In case of insufficient bandwidth, QoS helps to guaranteeburst dropping probability. QoS also helps to provide different priorities to differentapplications depending on their bandwidth requirement. So, recently QoS has gain muchattention and is becoming motivation for present research. In general, QoS can beprovided by introducing differentiation at some point in the network. Many algorithmshave been proposed to provide QoS in OBS based networks. In [2], a simple but costefficient priority scheme is proposed to support QoS at the WDM layer of the opticalInternet. More specifically, by assigning different offset times to different traffic classes,it is possible for a higher priority class to be isolated from lower priority traffic class.Though offset-time based QoS scheme does not mandate the use of Fiber Delay Lines(FDLs), in [3,4] it is given that QoS performance can be significantly improved even withlimited FDLs to resolve contention for bandwidth among multiple bursts. Theproportional QoS model in [10] is proposed which introduces an intentional droppingscheme to give a controllable burst loss probability for different service classes.Rescheduling based QoS control algorithm using delayed pre-emption and controlleddeadlines has been proposed in [14]. 219
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEMEII. SERVICE DIFFERENTIATION IN OBS There is a great demand for Internet to be extended with service differentiation. Aswe have said earlier, the QoS can be provided by introducing differentiation at some pointin the network. Various techniques of service differentiation are designed to provide QoSincluding differentiated signalling, differentiated offset-time [1], differentiated offset-time[2,3], differentiated contention resolution [8], differentiated burst assembly [6,7],differentiated scheduling [6], differentiated path selection [13]. Recently, a new approachis proposed in [12] which provide QoS based on differentiation scheme. This approachuses pre-emption for contention resolution and guarantees zero burst loss of high delaysensitive data by using independent wavelength assignment scheme for high delaysensitive data and synchronizing all core nodes by a single clock to avoid wavelengthcontention but it has some limitations. There is a need of synchronization of all corenodes to set lightpath. The additional mechanism is required to calculate number ofwavelengths for high delay sensitive data at each core node. This algorithm uses the staticwavelength assignment scheme for high delay sensitive data bursts. We are proposing a new approach which uses dropping technique for contentionresolution and also reduces the loss of high delay sensitive data. In the proposed approachno lightpath is created for high delay sensitive data transmission. It means that proposedalgorithm does not need independent wavelength assignment for high delay sensitive dataand clock synchronization at each core node.The additional mechanism is not required tocalculate number of wavelengths for high delay sensitive data at each core node. So, thecomplexity of proposed algorithm is reduced. The proposed algorithm uses packet leveldifferentiation in which packets from different service classes are assembled into differentbursts. In this paper, like [12] the Internet traffic at ingress node is classified into threeclasses of service: Most high priority data, High priority data and Low priority data. Thisclassification isbased on the current and next generation demands ofthe users and realtime applications which is the reasonof classifying video further in to live videos andstreaming videos. The concept of service differentiation has been employed to reduce theloss of MHP data bursts.III. PROPOSED APPROACH In order to describe the proposed dropping based contention resolution algorithm,it is assumed that wavelength converters and FDLs are used at each core node. TheInternet traffic at ingress node is classified into three classes of service: Most high prioritydata, High priority data and Low priority data. High delay sensitive data packets (e.g. livevideo conference, voice etc) are assembled into a bursts of most high priority (MHP), thedata packets (e.g. streaming video etc) are assembled into high priority (HP) bursts, whilebest effort data packets (e.g. FTP, email etc) are assembled into low priority (LP) bursts. All available wavelengths are divided MHP data channels, HP data channels andLP data channels. Proposed scheme uses dropping technique for wavelength contentionresolution. The idea is that if data burst having high priority is contending with burst withlow priority, then drop the low priority data to schedule data burst having high priority. 220
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEMEA. Algorithm In proposed approach, bursts are generated by using threshold-based scheme of burstaggregation. It is ensure that BHP is processed before data burst is transmitted. The timecomplexity of proposed algorithm is O(n). BEGIN Check the class of incoming burst IF MHP burst is at core IF λ is available Generate control packet and transmit burst after offset time ELSE Check the possibility to lease λ IF Possible to lease LP data channel Channel is reserved to transmit MHP burst after offset time ELSE IFPossible to lease HP data channel Channel is reserved to transmit MHP burst after offset time ELSE MHP burst is dropped END IF END IF END IFELSE IF HP burst is at core IFλ is available Generate control packet and transmit burst after offset time ELSE Check possibility to lease λ IFPossible to lease LP data channel Channel is reserved to transmit HP burst after offset time ELSE HP burst is dropped END IF END IF ELSELP burst is at core IFλ is available Generate control packet and transmit burst after offset time ELSE LP burst is dropped END IF END IF END IFEND 221
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEMEB. Operation Of Proposed Algorithm When any burst arrives at core node then algorithm first checks service class of the burst.If type of the burst is MHP then algorithm checks whether the wavelength is available toschedule arriving data burst. If wavelength is available then it is assigned, control packet isgenerated and MHP burst is scheduled on channel after offset time to transmit. If wavelengthis not available for MHP burst then algorithm first checks the possibility to lease LP datachannel. If it is possible then LP data channel is reserved to schedule MHP burst after offsettime. If it is not possible to lease LP data channel then again a check is made to lease HP datachannel. If it is possible to lease HP data channel then HP data channel is reserved toschedule MHP burst and transmitted after offset time else MHP data burst is dropped asshown in flowchart 1. It means that algorithm has given the full authority to MHP burst tolease the wavelengths of LP and HP bursts so that it helps to minimize dropping probabilityof MHP burst. So, proposed algorithm guarantees to reduce the loss of MHP burst. If the service class of burst at core is HP, then algorithm checks whether wavelength isavailable to schedule arriving burst. If wavelength is available then wavelength is assigned,control packet is generated and HP burst is scheduled on channel to transmit after offset time.If wavelength is not available for burst then algorithm checks the possibility to lease LP datachannel. If it is possible, then LP data channel is reserved to schedule HP burst after offsettime. If it is not possible to lease LP data channel, then HP burst is dropped shown inflowchart 3. Flowchart 1: Operation of Algorithm when MHP burst at core node If the class of arriving burst at core is LP, then algorithm checks whether wavelength isavailable to schedule arriving LP burst. If wavelength is available then wavelength isassigned, control packet is generated and LP burst is scheduled on channel to transmit afteroffset time. If wavelength is not available LP burst is dropped shown in flowchart 2. 222
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976- 6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEMEFlowchart 2: Operation of Algorithm when LP Flowchart 3: Operation of Algorithm when HP burst at burst at core node core node The dropping probability of HP class data burst depends on the load of MHP data burst. As number of MHP data burst increases, possibility of HP burst drop also increases. Similarly, the dropping probability of LP class data burst depends on the percentage load of both MHP and HP bursts. As number of MHP and HP data burst increases, possibility of dropping of LP burst also increases. C. Wavelength Allocation Scheme The wavelength allocation scheme in the proposed algorithm is not static. The total available wavelengths are grouped as MHP data channels, HP data channels, and LP data channels. To make wavelength allocation dynamic, MHP data bursts are allowed to lease wavelengths of HP and LP data channels, while HP data bursts are allowed to lease LP data channels if wavelength is not available. But, this scheme may lead to total loss of LP data burst. So, a count is placed at each core node to measure number of LP burst drop and is not allowed if count reaches to a threshold value. Similarly, another count is placed at each core node to measure HP burst drop and if count reaches to its threshold value, then dropping of HP burst is not allowed. Figure 1 shows the dynamic nature of wavelength allocation scheme. IV. SIMULATION AND RESULTS Extensive simulations are carried out to study the performance of proposed dropping based approach and to see how it fares in comparison with standard LAUC-VF algorithm. To evaluate the performance of proposed algorithm NS-2.33 with OBS patch is used. In this section, the method used for conducting simulations is described and the performance parameter used to analyse the proposed approach is defined. Finally, the results obtained from the simulation are presented. 223
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEME Figure 1: Wavelength Allocation Scheme in Proposed AlgorithmA. Simulation Model We have conducted simulations using NS-2.33 integrated with OBS patch andNSFNET topology consisting of total 14 nodes. The simulation model consists of 10 corenodes and 4 edge nodes with 21 bidirectional links. A bidirectional link is realized by twounidirectional links in opposite direction. Each unidirectional link consists of 7 datachannels and 1 control channel as shown in figure 2. Burst arrivals to the network areSelf-similar. Bursts are generated by using Threshold-based scheme. Shortest path isused for routing the burst from source to destination. The BHP processing time is 1 µs. Itis ensure that BHP is processed before data burst is transmitted.B. Performance Metric The burst dropping probability metric is selected to study the performance of proposedalgorithm. The burst dropping probability is defined as number of bursts dropped overnumber bursts sent. The efficiency of scheduling algorithm depends on burst droppingprobability and it should be as minimum as possible. Because, burst dropping probabilityis very important factor as it leads to very huge amount of data loss in optical network. Load in the network depends on total bandwidth available and burst size. The maximumload that can be handled by network is calculated by dividing total available wavelengthby burst size. 224
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEME Figure 2: Simulation ModelC. Simulation Results The performanceof proposedalgorithm in terms of dropping probability and compares itwith LAUC-VF algorithm is shown in figure 3. Proposed algorithm shows the reduction inloss of MHP class data bursts. Even though data load is 100%, the burst loss recorded isminimum than LAUC-VF, because MHP burst can lease channels which are allocated to LPand HP data channels. It has been observed that using LAUC-VF when data load is 100%,dropping probability reaches to 0.14 [12]. Similarly, figure 4 shows effect of data load on dropping probability of HP bursts. Itclearly shows that as data load increases, dropping probability of HP also increases.Because, as data load increases MHP load also increases and MHP burst can lease channelswhich are allocated to HP and those which can be leased by HP. Figure 3: Comparison of LAUC-VF and Proposed Algorithm 225
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEME Figure 4: Behaviour of Dropping Probability of HP Bursts with Increasing Load of MHP Bursts The effect of data load on dropping probability of LP bursts is shown in figure 5. It isobserved that as data load increases, dropping probability of LP data also increases. Withincreasing load of MHP class data bursts, the availability of wavelengths decreases. In thatcase, MHP burst is going to lease channels of LP to reduce dropping. As a result, there isincrease in dropping of LP. Figure 5: Behaviour of Dropping Probability of LP Bursts with Increasing Load of MHP Bursts 226
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEMEV. CONCLUSION In this paper, a simple and effective dropping based scheme is proposed which usespacket level service differentiation by dividing the IP traffic into three service classes. It hasbeen observed that proposed algorithm reduces the loss of high delay sensitive data. The ideais that if wavelength is not available to schedule burst then wavelength of burst having lowpriority can be can be dropped to lease that wavelength. MHP burst loss is reduced, becauseMHP class data burst can lease the wavelengths of both LP and HP class data burst. To avoidtotal loss of LP burst, some threshold value is placed at each core node to measure LP drop.Simulations are carried out to study the performance of proposed approach and comparedwith standard LAUC-VF algorithm. While implementing proposed algorithm, it is observedthat as MHP class data burst load increases, dropping probability of LP and HP data burstalso increases.REFERENCES[1] Jason P. Jue, Vinod M. Vokkarane. “Optical Burst Switched Networks”, Springer Publication, 2005.[2] MyungsikYoo and ChunmingQiao. ``Supporting Multiple Classes of Services In IP Over WDM Networks, IEEE CONFERENCE, Page(s) 1023 - 1027 vol. 1b, 1999.[3] MyungsikYoo, ChunmingQiao and Sudhir Dixit. ``QoS Performance of Optical Burst Switching in IP-Over-WDM Networks, IEEE JOURNALS, Page(s) 2062 - 2071, 2000.[4] MyungsikYoo, ChunmingQiao and Sudhir Dixit. ``Optical Burst Switching for Service Differentiation in the Next-Generation Optical Internet, IEEE Communications Magazine, Page(s) 98 - 104, 2001.[5] Mei Yang, S.Q. Zheng and Dominique Verchere. ``A QoS Supporting Scheduling Algorithm For OBS DWDM Networks, IEEE Global Telecommunications Conference, Page(s) 86 - 91 vol.1, 2001.[6] MyungsikYoo, ChunmingQiao and Sudhir Dixit. “Optical Burst Switching for Service Differentiation in the Next-Generation Optical Internet IEEE Communications Magazine, Page(s) 98 - 104, 2001.[7] Vinod M. Vokkarane, Qiong Zhang, Jason P. Jue, and Biao Chen. “Generalized Burst Assembly and Scheduling Techniques for QoS Support in Optical Burst-switched Networks IEEE Global Telecommunications Conference, Volume 3, Page(s) 2747 - 2751 vol.3, 2002.[8] Ching-Fang Hsu, Te-Lung Liu and Nen-Fu Huang. “On the Deflection Routing in QoS Supported Optical Burst-Switched Networks Communications, 2002. ICC 2002. IEEE International Conference, Volume 5, Page(s) 2786 - 2790, 2002.[9] TzvetelinaBattestilli and Harry Perros, “An Introduction to Optical Burst Switching” IEEE Optical Communications August 2003[10] YunhanLuo, Sheng Wang. “An FDL-based QoS scheduling algorithm in OBS Networks IEEE, 2005.[11] M. Casoni, E. Luppi and M. L. Merani. “Performance Evaluation of Channel Scheduling Algorithm With Different QoS ClassesNetworks, 2006. ICON 06. 14th IEEE International Conference, Volume 2, Page(s) 1 - 6, 2006. 227
  • International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 – 6375(Online) Volume 4, Issue 1, January- February (2013), © IAEME[12] FarhanSabir Ujager1, Zia-ul-Haq2, Muhammad Ramzan3 and S.M.H. Zaidi.“An independent wavelength assignment, QoS Differentiation Scheme, in DWDM OBS Networks, for zero high priority burst loss" IEEE CONFERENCE, Page(s) 274 - 279, 2010.[13] M.Thachayani, Member IEEE and R.Nakkeeran. “Path Differentiated QoS Provisioning Scheme for OBS Networks, IEEE 2010 International Conference, Page(s) 1 - 4, 2010.[14] Kostas Ramantas, Tito Raul Vargas, Juan Carlos Guerri and Kyriakos Vlachos. “A preemptive scheduling scheme for flexible QoS provisioning in OBS networks IEEE Sixth International Conference, Page(s) 1 - 6, 2009.[15] V. Bapuji, R. Naveen Kumar, Dr. A. Govardhan and Prof. S.S.V.N. Sarma, “Maximizing Lifespan Of Mobile Ad Hoc Networks With Qos Provision Routing Protocol” International journal of Computer Engineering & Technology (IJCET), Volume 3, Issue 2, 2012, pp. 150 - 156, Published by IAEME.[16] Jayashree Agrakhed, G. S. Biradar and V. D. Mytri, “Optimal QoS Routing With Prioritized Region Scheduling Over WMSN” International journal of Computer Engineering & Technology (IJCET), Volume 3, Issue 1, 2012, pp. 289 - 304, Published by IAEME. 228